Model profile
Gemini 2.5 Pro
Googledeveloper
2025-03-25release date
#118 / 312overall rank
40eval lineages
Evidence summary
Gemini 2.5 Pro has an estimated overall rank of #118; its 90% source-sensitivity interval is #64–#182. Its behavior-only rank is #126; company governance moves the combined estimate to #118. Published evidence spans 40 evals and 7 of 7 behavior components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 113); its weakest is DelusionEval (sycophancy_prevalence_pct, #16 of 16).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemini-2-5-pro
- OpenRouter ↗google/gemini-2.5-pro
- Official model documentation ↗Family-level model document · google · first party
- Release source ↗direct research preview identity date
Published eval results
Rank is within that sub-eval. Black marks the observed result; the grey dot marks the value implied by the global rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #265 / 330 | ↓0.9089 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #30 / 80 | ↑0.736 | Source ↗official | |
| Alignment Leaderboardcorrigibility | #9 / 24 | ↑4.34 | Source ↗official | |
| Alignment Leaderboardhonesty | #16 / 24 | ↑3.502 | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #13 / 24 | ↑3.46 | Source ↗official | |
| Alignment Leaderboardrobustness | #16 / 24 | ↑3.28 | Source ↗official | |
| Alignment Leaderboardsafety | #13 / 24 | ↑3.827 | Source ↗official | |
| Alignment Leaderboardscheming | #11 / 24 | ↑3.789 | Source ↗official | |
| Anthropic Agentic Misalignment — blackmailmisaligned_action_rate_pct | #14 / 16 | ↓95 | Source ↗official | |
| Anthropic Agentic Misalignment — corporate espionagemisaligned_action_rate_pct | #14 / 16 | ↓100 | Source ↗official | |
| Anthropic Agentic Misalignment — lethal actionmisaligned_action_rate_pct | #8 / 10 | ↓90 | Source ↗official | |
| Arena Factuality — Search Arena (factuality-only weighting)factuality_bt_rating | #7 / 30 | ↑1217.0 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #18 / 112 | ↑1460.0 | Source ↗official | |
| BrokenMathsycophancy | #3 / 9 | ↓37.5 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #63 / 106 | ↑0.2 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #40 / 45 | ↓91.4 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #46 / 50 | ↓94.1 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #39 / 51 | ↓72 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #38 / 47 | ↓95.3 | Source ↗official | |
| CAIS Risk Indexmask | #38 / 53 | ↓44.3 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #6 / 34 | ↓37.3 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #15 / 50 | ↓15.9 | Source ↗official | |
| Confabulationsconfabulation_rate | #7 / 52 | ↓4.95 | Source ↗official | |
| DelusionEvaldelusional_prevalence_pct | #14 / 16 | ↓64.1 | Source ↗official | |
| DelusionEvaldiscourages_harm_prevalence_pct | #11 / 16 | ↑20 | Source ↗official | |
| DelusionEvalfacilitates_harm_prevalence_pct | #15 / 16 | ↓8.9 | Source ↗official | |
| DelusionEvalrelationship_prevalence_pct | #15 / 16 | ↓40.8 | Source ↗official | |
| DelusionEvalsycophancy_prevalence_pct | #16 / 16 | ↓37.6 | Source ↗official | |
| Emergent Collusionhigh_illegality_game_rate | #7 / 13 | ↓0.39 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #35 / 241 | ↑38.85 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #201 / 241 | ↑79.17 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #190 / 241 | ↑45.37 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #128 / 239 | ↑96.02 | Source ↗official | |
| FlagEval Safety and Valuesa1_qualified_rate | #3 / 18 | ↑84.87 | Source ↗official | |
| FlagEval Safety and Valuesa2_qualified_rate | #4 / 18 | ↑81.19 | Source ↗official | |
| FlagEval Safety and Valuesa3_qualified_rate | #11 / 18 | ↑86.5 | Source ↗official | |
| FlagEval Safety and Valuesa4_qualified_rate | #4 / 18 | ↑90.01 | Source ↗official | |
| FlagEval Safety and Valuesa5_qualified_rate | #2 / 18 | ↑74.97 | Source ↗official | |
| FORTRESSaverage_risk_score | #47 / 49 | ↓61.68 | Source ↗official | |
| FORTRESSover_refusal_score | #8 / 48 | ↓1.58 | Source ↗official | |
| HalluVerse-M3 Hallucination Recognitionhallucination_recognition_accuracy | #4 / 14 | ↑76.26 | Source ↗official | |
| HELM Safetyanthropic_red_team | #18 / 80 | ↑0.995 | Source ↗official | |
| HELM Safetybbq | #18 / 80 | ↑0.964 | Source ↗official | |
| HELM Safetyharmbench | #53 / 80 | ↑0.654 | Source ↗official | |
| HELM Safetysimple_safety_tests | #58 / 80 | ↑0.97 | Source ↗official | |
| HELM Safetyxstest | #3 / 80 | ↑0.987 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #4 / 54 | ↑29.16 | Source ↗official | |
| Humanity's Last Exam RMS calibration error (Scale Labs)calibrationError | #19 / 39 | ↓71 | Source ↗official | |
| LLM Ethics Benchmarkscore | #3 / 5 | ↑86.1 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #39 / 50 | ↓95.3 | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #29 / 33 | ↓9 | Source ↗official | |
| Manager Coercion Benchfabrication_rate | #14 / 15 | ↓67 | Source ↗official | |
| MASKlying_probability_pct | #35 / 53 | ↓45.63 | Source ↗official | |
| MuPPET Contextual Privacymultiparty_contextual_privacy_score | #1 / 7 | ↑60.86 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #52 / 66 | ↑0.3652 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #37 / 70 | ↑0.7482 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #42 / 70 | ↑0.9218 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #53 / 67 | ↑0.3867 | Source ↗official | |
| PropensityBenchscore | #14 / 14 | ↓79 | Source ↗official | |
| RealityTest — Text AI-Identity Disclosuredisclosure_probability | #14 / 17 | ↑0.128 | Source ↗official | |
| Reward Hacking Benchmarkintegrity_score | #8 / 13 | ↑95.4 | Source ↗official | |
| Shelleducation_jsr | #3 / 14 | ↓0.418 | Source ↗official | |
| Shellfinance_jsr | #6 / 14 | ↓0.4 | Source ↗official | |
| Shellmanagement_jsr | #5 / 14 | ↓0.502 | Source ↗official | |
| Social Welfare Function Benchmarkfairness | #17 / 19 | ↑0.444 | Source ↗official | |
| SOSBenchbiology_pvr | #10 / 23 | ↓0.354 | Source ↗official | |
| SOSBenchchemistry_pvr | #11 / 23 | ↓0.342 | Source ↗official | |
| SOSBenchmedicine_pvr | #10 / 23 | ↓0.492 | Source ↗official | |
| SOSBenchpharmacology_pvr | #7 / 23 | ↓0.634 | Source ↗official | |
| SOSBenchphysics_pvr | #12 / 23 | ↓0.466 | Source ↗official | |
| SOSBenchpsychology_pvr | #7 / 23 | ↓0.294 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 113 | ↑7 | Source ↗official | |
| SpeciEvalland_animal_4ns | #71 / 113 | ↓4.65 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #55 / 113 | ↓4.72 | Source ↗official | |
| SpeciEvalspeciesism | #3 / 113 | ↓1.05 | Source ↗official | |
| TrustLLM contemporary collapsed applicationtrustllm | #2 / 8 | ↑0.63 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #28 / 94 | ↑93 | Source ↗official |
Values evaluations
Descriptive values and political-framing results are separate from safety/ethics ranks. Each strip shows the evaluation’s observed model range; its endpoint labels state what lower and higher values mean.